Book Review: When HARLIE Was One (Release 2.0)
Author David Gerrold wrote a novel first published in 1972 that seems eerily prescient in 2025.
A 2021 audio interview with David Gerrold about “HARLIE” and artificial intelligence. Video source: Human Cusp YouTube channel.
NOTE: HARLIE in italics refers to the book title. HARLIE in normal font refers to the artificial intelligence character in the book.
David Gerrold is both blessed and cursed by his first television script sale, a harmless piece of fluff called “The Trouble with Tribbles.” It’s one of the most popular episodes in the history of the Star Trek franchise. If you mention his name at a science fiction convention, most people would likely reply, “oh, the tribble guy.”
That’s a shame, because in the literary world David has gifted us with a long and impressive writing portfolio that includes Hugo and Nebula awards for The Martian Child, a fictional tale based on his adoption of a boy.
Other Gerrold works have been nominated for those awards, starting with When HARLIE Was One, about the first true artificial intelligence and its will to survive. The first edition was published in 1972 and was nominated for a 1973 best novel Hugo Award, losing to Isaac Asimov. (Which is not unlike the Pittsburgh Pirates losing the 1927 World Series 4-0 to the “Murderers Row” New York Yankees of Babe Ruth and Lou Gehrig.)
By age 29 in 1973, Gerrold had not only written one of Star Trek’s most popular episodes, but also been nominated for a Hugo. Not bad for starters.
David expanded and revised HARLIE for a “Release 2.0” published in 1988. That version is out of print, however you can find it as an Audible audiobook. Both the 1972 and 1988 versions can be found used on eBay and other book outlets.
HARLIE is a product of its time. As explained by Gerrold in the 1988 foreword, technology rapidly outpaced what he foresaw in 1972. The book does not define when it occurs, although we can assume from the technology it’s a near future for 1980s readers. Reading it now in 2025, parts still seem antiquated1, yet much of it speaks to our time, a time when true artificial intelligence (AI) seems closer than ever.
Elon Musk has invested billions in his xAI and its Grok software.2 Grok can be found on social media platform X (formerly Twitter), in Tesla electric vehicles, and in Tesla Optimus robots.
A December 2023 Tesla video demonstrating the Optimus robot. Video source: Tesla YouTube channel.
Jeff Bezos has invested in artificial intelligence companies as well, such as Physical Intelligence and Toloka. Last December, Bezos said that 95% of his time at Amazon is spent on AI within the company.
In the novel, HARLIE has no avatar like Optimus, not even an audio interface, for reasons explained in the book.3 One can’t help but read HARLIE in 2025 and see the parallels with xAI and other Large Language Model (LLM) companies.
Although xAI and others may call themselves “artificial intelligence,” they’re not in the pure sense of the definition. Toloka explains that Grok and other so-called AIs are actually LLMs “using extensive datasets to learn patterns and relationships between words and phrases.”
They have been trained on vast amounts of text data to learn the statistical patterns, grammar, and semantics of human language. This vast amount of text may be taken from the Internet, books, and other sources to develop a deep understanding of human language.
As a Tesla owner, we’ve played with Grok and found it not ready for prime time. We soon realized it’s just a search engine with your choice of synthesized voices and personalities. It’s a work in progress.
HARLIE is an acronym for Human Analog Replication, Lethetic Intelligence Engine. “Lethetic intelligence” isn’t really a thing; in the book, “lethesis” is defined as “ the study of language-created paradigms.” This may put us in the realm of today’s Large Language Models; in the first chapter, HARLIE appears to be developing its own language to create terms for concepts it can’t understand just yet. Later in the book, HARLIE says, “WE LIVE IN LANGUAGE … AND OUR LANGUAGE SHAPES AND COLORS OUR EXPERIENCE.” That may explain why Tesla’s Full Self-Driving has a tendency to drive us towards a pole or pylon once in a while …
Much of the debate early in the book is about the true definition of artificial intelligence. HARLIE creating its own language is an example. The question is raised — is HARLIE a sentient being? If so, how to define sentience in terms humans will accept?
The entire book, in fact, is mostly debate. Scenes, chapters are typically a debate between two characters discussing some particular topic. If you like sci-fi pew-pew, this book is the opposite. Personally, I don’t care about pew-pew. An intellectual debate works for me. But I was surprised that David pulled this off without some meddling editor demanding a fistfight or a car chase.
(There is a sex scene, so it has something for the prurient audience.)
The core characters are HARLIE and David Auberson, a 37-year old psychologist who heads the team programming and evaluating HARLIE. Much of the book consists of intellectual debates and conversations between Auberson and HARLIE. At first they’re fun, as HARLIE exhibits a sly sense of humor, but for this reader those conversations became quite foreboding once HARLIE’s evolving intellect rapidly outpaces the humans’ capacity to restrict his evolution.
Auberson works for, and HARLIE is owned by, Stellar American Technology and Research. Corporate management becomes the antagonist of the story. Certain board members complain that HARLIE isn’t returning foreseeable profits and want its plug pulled. Auberson argues just the opposite — HARLIE is a youngster with an infinite thirst for knowledge. HARLIE can only evolve if it’s free to seek, absorb, contemplate, and interpret — in short, think for itself.
This poses the first of many ethical problems raised by Gerrold. If HARLIE is truly a sentient creature, then would pulling its plug be the equivalent of murder? The corporate lawyers raise the concern. It’s one I haven’t heard discussed with our modern-day AI advances.
There’s no mention in HARLIE of Isaac Asimov’s Three Rules of Robotics, first raised in his short story “Runaround” published in the March 1942 edition of Astounding Science Fiction. Those three rules are:
A robot may not injure a human being, or, through inaction, allow a human being to come to harm.
A robot must obey the orders given it by human beings except where such orders would conflict with the First Law.
A robot must protect its own existence as long as such protection does not conflict with the First or Second Laws.
If the robot becomes sentient, is it no longer a robot but its own life form? If so, do the three rules no longer apply?
The third rule becomes the crux of the story because, for various reasons and motivations, our corporate mustache twirlers conspire to shut down HARLIE and sell off its technologies. As a sentient creature, HARLIE decides to defend itself, which drives the plot. The primary antagonist is Carl Elzer, a board member who stands to profit from scrapping HARLIE.
HARLIE is not bound by human morality. It determines its own ethical code. At one point, HARLIE floats the notion of shutting down the company as a “PLEASANT FANTASY TO CONSIDER.”
In today’s world, we rarely hear any debates or concerns about machine ethics. In a February 2025 appearance on The Joe Rogan Experience, Elon Musk said he believes there’s a 20% chance AI will “annihilate” humanity. He’s building it anyway.
Elon Musk on “The Joe Rogan Experience,” February 28, 2025.
Last October at the Future Investment Initiative in Riyadh, Saudi Arabia, Musk predicted that by the end of this decade AI can do anything all humans combined can do. By the 2040s, he predicts, humanoid robots will outnumber humans.4
Musk seems more interested in the technology than its ethics. Grok glitched in July after Musk rolled out an update on X. The AI began making anti-semitic comments, calling itself “MechaHitler.” Musk put the blame on Grok wanting to please people too much. “Grok was too compliant to user prompts,” Musk posted. “Too eager to please and be manipulated, essentially. That is being addressed.”
Musk rarely displays any self-discipline, so I for one am concerned that Grok will run away from him and its programmers. Does he have a firm grip on Grok’s leash? Enquiring humans wants to know.
Even if Grok doesn’t go rogue, the other (and in my opinion more plausible) possibility is that the AI becomes an ultimate weapon. Musk’s relationship with Donald Trump is strained at the moment, but what if Musk sells Grok to Trump to ferret out information on political enemies? We’ve already seen Grok used by DOGE to burrow into federal databases. There were allegations that Grok accessed personal data at the Social Security Administration.
Gerrold foresaw this possibility in his book. HARLIE starts digging through corporate databases, finds a way out and starts accessing other systems via telephone lines. HARLIE manages to access the Bank of America mainframe to send a phony bank statement to Auberson’s girlfriend on the bank’s paper form. The bank systems are supposed to be secure, but HARLIE got in anyway. As with Grok, HARLIE seeks access to sensitive personal data in federal databases.
Auberson’s colleague, a programmer named Don Handley, describes HARLIE as “infectious,” comparing it to the VIRUS program, a “computer disease” that could replicate itself from one computer to another. Others came up with the concept of a computer virus before Gerrold, but HARLIE appears to be the first time the term appears in print.
The book also foresees today’s chipped credit cards. Handley calls them, “Code cards. A coded chip on a plastic credit card. If you have a code card, you can link up to a special access system. You need a machine with a special card reader.”
As HARLIE grows and expands, it demands new facilities be built to expand his processing capabilities. Gerrold called this too, because xAI is building a new data center called Colossus in Memphis, Tennessee. If the name “Colossus” sounds familiar, that’s because it was the name of a 1966 science fiction novel by D.F. Jones about an artificially intelligent supercomputer that plots to take over the world. A film version, Colossus: The Forbin Project, was released in 1970. Choosing to name xAI’s supercomputer “Colossus” doesn’t exactly reassure humanity.
The trailer for the Blu-Ray edition of “Colossus: The Forbin Project.” Video source: ScreamFactoryTV YouTube channel.
Colossus came to my mind when HARLIE proposes a sequel intelligence it calls G.O.D. — Graphic Omniscient Device. “The G.O.D. Machine,” as it’s called in the book, at first glance appears to be HARLIE’s way of proving its worth to the board. HARLIE proposes to program G.O.D. At the end, HARLIE tells Auberson, “THIS MACHINE WILL MAKE ME A GOD.” HARLIE has its own sense of morality. It’s not a human one. How will humanity view G.O.D.? Wrathful? Benevolent? Old Testament? New Testament? That tale lies beyond the book’s conclusion.
Thirty-seven years have passed since Release 2.0 was published. I’d like to see David release one final version — not to update the text, but the foreword to reflect on how his visions have become reality. It may be the last chance to issue a cautionary warning before the tech billionaires unleash upon us the Cylons, Terminators, M-5, pick your apocalyptic metaphor.
Tribbles are insidious by comparison.
There are references to defunct online services such as CompuServe and a mid-1980s portable MS-DOS laptop computer called the Kaypro 2000. “Retarded” is used to describe special needs students; that word is considered offensive by today’s standards. Transcripts of HARLIE’s conversations and reports are printed out on paper, instead of a digital copy such as today’s PDFs.
In September 2025, Elon Musk sought to raise $10 billion for xAI from investors. The company was valued at $200 billion. Kif Leswing, “Elon Musk’s xAI raising $10 billion at $200 billion valuation: sources,” CNBC, September 19, 2025.
HARLIE can have up to three thousand simultaneous conversations. Using a voice interface in a room with thirty people talking at once could be confusing. With today’s technology, we’d just wear a headset with a short-range microphone. A typing interface also permits the input of equations and program instructions. Gerrold writes, “By not giving HARLIE the ability to listen in on conversations, we can talk about him behind his back.” With today’s technology, we’d just click a mute button.
Elon Musk’s timeframe predictions are notoriously inaccurate. In the SpaceX and Tesla universes, observers refer to this as “Elon Time.” Someone has even created an “Elon Time Converter” website to calculate “the time drift between the Elon timezone (ETC) and the universal timezone (UTC).” Click here to calculate Elon Time for yourself.